Application of multiple regression analysis to the prices of the spectrum in the IMT band

This paper explores the effects of different variables on the price paid for the spectrum by Mobile Service Providers (MSP) in several European countries. Multiple regression analysis is used to find relations between these variables and the Price/MHz/pop, commonly used in the industry to compare prices across bands and markets. The results show that the most important factor that affects spectrum prices is the frequency band, followed by the country, and, to a lesser extent, the number of competitors and the year of allocation. The analysis is also used to create a regression function capable of predicting future prices. The model has been cross-validated with data from past tenders, with mixed results. The lowest prediction errors are around 10%. Even if accuracy is not high, the model provides a valid range of price estimates in which real prices oscillate. Further research is needed regarding variables not being currently captured in the model, as the balance between supply and demand of spectrum at any one awarding process.